Tumor detection method and system based on image analysis

Abstract

The invention discloses a tumor detection method and system based on image analysis, and relates to the technical field of image data analysis. The tumor detection method based on image analysis comprises the following steps: acquiring lung image data of a patient to be detected, preprocessing the lung image data, inputting the lung image data into a pre-trained region recognition model for recognition and analysis to obtain a plurality of tumor regions of the patient to be detected, and comprehensively analyzing the tumor regions to obtain tumor staging indexes of the patient to be detected; meanwhile, physiological data of the patient to be detected are obtained, data analysis is carried out, the tumor physiological load index of the patient to be detected is obtained, comprehensive analysis is carried out in combination with the tumor staging index, and the tumor progress index of the patient to be detected is obtained. And corresponding treatment measures are taken based on detection analysis, so that the accuracy of tumor detection is improved, the phenomenon of excessive treatment is effectively avoided, the risk of delayed treatment is further avoided, and the survival rate of a patient is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image data analysis, and particularly to a tumor detection method and system based on image analysis. Background Art

[0002] Image analysis technology, especially its application in the field of medical imaging, has become an important means for disease detection and treatment plan formulation. Image analysis involves extracting effective information from medical images and converting it into data for decision-making support. For example, the detection of lung tumors, especially lung cancer, is one of the main diseases causing death globally. Early detection and timely diagnosis of lung cancer are crucial for improving survival rates. The treatment effect and survival period of patients with lung cancer are closely related to the diagnostic stage of the tumor. Due to the lack of obvious symptoms in early-stage lung cancer, many patients are often diagnosed when the tumor has spread or metastasized. Therefore, how to accurately evaluate the diagnostic stage of tumors through imaging technology is an important topic in current medical research and clinical applications.

[0003] A prior art tumor detection method disclosed in a patent application with publication number CN112288672B specifically includes the following steps: obtaining a detection image; cutting the detection image into multiple image blocks according to the input size of the training data of the convolutional neural network architecture, and marking the coordinate values of each image block before cutting; respectively inputting the multiple image blocks into a preset recognition and detection model to obtain corresponding detected result image blocks; and splicing the detected result image blocks according to the coordinate values of the image blocks to form a detected result image, so as to obtain the detection result of the tumor. The present invention also provides a tumor detection device. The present invention can detect whether there is a tumor through images and determine the specific location of the tumor, with a novel concept, fast detection speed, and high detection accuracy.

[0004] Based on the above solution, it is found that the limitations of the prior art at least include the following problems. In the process of tumor detection in the prior art, it is difficult to retain tumor boundary information, especially in the case of complex tumor morphology and blurred edges, resulting in inaccurate recognition of the tumor area and lack of comprehensive analysis of the patient's physiological data, thus making it easy for the detection result to be difficult to comprehensively reflect the tumor progression of the patient, especially showing a lag in the evaluation of advanced tumors. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides a tumor detection method and system based on image analysis, which solves the problems that in the process of tumor detection in the prior art, it is difficult to retain tumor boundary information and lack of comprehensive analysis of the patient's physiological data.

[0006] To achieve the above object, the present invention is realized through the following technical solutions: A tumor detection method based on image analysis, comprising the following steps: acquiring lung image data of a patient to be detected and performing preprocessing; inputting the preprocessed lung image data of the patient to be detected into a pre-trained region recognition model for recognition and analysis to obtain several tumor regions of the patient to be detected, and performing comprehensive analysis to obtain a tumor staging index of the patient to be detected; simultaneously acquiring physiological data of the patient to be detected and performing data analysis to obtain a tumor physiological load index of the patient to be detected, and combining with the tumor staging index for comprehensive analysis to obtain a tumor progression index of the patient to be detected; performing detection and analysis based on the tumor progression index of the patient to be detected, and taking corresponding treatment measures based on the detection and analysis; wherein, the specific formula for calculating the tumor progression index of the patient to be detected is as follows: Wherein, FzL, ZqF, and ShF are the tumor progression index, tumor staging index, and tumor physiological load index of the patient to be detected in sequence, ɑ1, β1, ɑ2, β2, and β3 are the staging coefficient, staging adjustment coefficient, load coefficient, load adjustment coefficient, and interaction coefficient stored in the database in sequence, and α1 + α2 = 1.

[0007] Further, the lung image data is specifically the pixel value and two-dimensional coordinate of each pixel point, and the physiological data includes hemoglobin concentration value, plasma lactate level value, tumor marker index, serum electrolyte level index, oxygenation index, and endocrine index.

[0008] Further, the region recognition model is specifically a U-shaped network, and the U-shaped network includes an encoder, a bottleneck layer, a decoder, a skip connection layer, and an output layer. The specific steps for obtaining several tumor regions of the patient to be detected are as follows: in the encoder of the U-shaped network, feature extraction is performed on the pixel value and two-dimensional coordinate of each pixel point of the patient to be detected to obtain a low-resolution tumor feature map of the patient to be detected; in the bottleneck layer of the U-shaped network, compression processing is performed on the low-resolution tumor feature map of the patient to be detected to obtain a high-dimensional tumor feature map of the patient to be detected; in the decoder of the U-shaped network, restoration processing is performed on the high-dimensional tumor feature map of the patient to be detected to obtain a restored tumor feature map of the patient to be detected; in the skip connection layer of the U-shaped network, fusion processing is performed on the restored tumor feature map of the patient to be detected to obtain a fused tumor feature map of the patient to be detected; in the output layer of the U-shaped network, recognition processing is performed on the fused tumor feature map of the patient to be detected to obtain several tumor regions of the patient to be detected.

[0009] Further, the specific steps to obtain the tumor staging index of the patient to be detected are as follows: Read the two-dimensional coordinates of each pixel point in each tumor region of the patient to be detected, perform median analysis to obtain the two-dimensional coordinates of each tumor region of the patient to be detected, and perform comprehensive analysis to obtain the tumor distribution index of the patient to be detected; perform comprehensive analysis on the two-dimensional coordinates of each pixel point in each tumor region of the patient to be detected to obtain the smoothness index of each tumor region of the patient to be detected; and perform gray-scale processing on the pixel values of each pixel point in each tumor region of the patient to be detected to obtain the gray-scale pixel value of each pixel point in each tumor region of the patient to be detected, and perform comprehensive analysis with the gray-scale pixel values of each neighborhood pixel point in the set neighborhood respectively to obtain the texture index of each tumor region of the patient to be detected; and perform comprehensive analysis on the tumor distribution index of the patient to be detected and the smoothness index and texture index of each tumor region to obtain the tumor staging index of the patient to be detected.

[0010] Further, the specific formulas for calculating the texture index and tumor staging index of each tumor region of the patient to be detected are as follows: Among them, WzS i is the texture index of the i-th tumor region of the patient to be detected, Xa ijm is the gray-scale pixel value of the m-th neighborhood pixel point in the set neighborhood of the j-th pixel point in the i-th tumor region of the patient to be detected, Xa ij is the gray-scale pixel value of the j-th pixel point in the i-th tumor region of the patient to be detected, S(X) is the sign function, ZqF and QbF are the tumor staging index and tumor distribution index of the patient to be detected in sequence, PhS i is the smoothness index of the i-th tumor region of the patient to be detected, δ1, δ2, δ3, and δ4 are the distribution adjustment coefficient, smoothness adjustment coefficient, texture adjustment coefficient, and superposition coefficient stored in the database in sequence, i = 1, 2, 3,..., i0, i0 is the number of tumor regions, j = 1, 2, 3,..., j0, j0 is the number of pixel points, m = 1, 2, 3,..., m0, m0 is the number of neighborhood pixel points, and e is the natural constant.

[0011] Further, the specific steps to obtain the smoothness index of each tumor region of the patient to be detected are as follows: Read the two-dimensional coordinates of each boundary pixel point of each tumor region of the patient to be detected, and perform comprehensive analysis in sequence to obtain the curvature index and boundary uniformity index of each tumor region of the patient to be detected; Read each pixel point of each tumor region of the patient to be detected, and perform comprehensive analysis to obtain the actual area value of each tumor region of the patient to be detected; And perform convex hull analysis on the two-dimensional coordinates of each boundary pixel point of each tumor region of the patient to be detected to obtain the convex hull area value of each tumor region of the patient to be detected, and perform comprehensive analysis in combination with the actual area value to obtain the compactness index of each tumor region of the patient to be detected; And perform comprehensive analysis on the curvature index, boundary uniformity index, and compactness index of each tumor region of the patient to be detected to obtain the smoothness index of each tumor region of the patient to be detected.

[0012] Further, the specific steps to obtain the tumor physiological load index of the patient to be detected are as follows: Obtain the hemoglobin concentration reference value, plasma lactate level reference value, tumor marker reference index, serum electrolyte level reference index, oxygenation reference index, and endocrine reference index of the patient to be detected; Perform comprehensive analysis on the hemoglobin concentration reference value, plasma lactate level reference value, tumor marker reference index, oxygenation reference index, hemoglobin concentration value, plasma lactate level value, tumor marker index, and oxygenation index of the patient to be detected to obtain the tumor physiological function influence index of the patient to be detected; And perform comprehensive analysis on the serum electrolyte level reference index, endocrine reference index, serum electrolyte level index, and endocrine index of the patient to be detected to obtain the physiological regulation index of the patient to be detected; And perform comprehensive analysis on the tumor physiological function influence index and physiological regulation index of the patient to be detected to obtain the tumor physiological load index of the patient to be detected.

[0013] Further, the specific formulas for calculating the tumor physiological function influence index, physiological regulation index, and tumor physiological load index of the patient to be detected are as follows: Among them, LsG is the tumor physiological function impact index of the patient to be detected. CkH, CsP, CbW, and CkY are the reference values of hemoglobin concentration, reference value of plasma lactate level, reference index of tumor markers, and reference index of oxygenation of the patient to be detected in sequence. XhD, RsP, ZbW, and YhZ are the hemoglobin concentration value, plasma lactate level value, tumor marker index, and oxygenation index of the patient to be detected in sequence. μ1, μ2, μ3, μ4, and μ5 are the hemoglobin adjustment coefficient, lactate adjustment coefficient, marker adjustment coefficient, oxygenation adjustment coefficient, and function impact superposition coefficient stored in the database in sequence. TsL is the physiological regulation index of the patient to be detected. XdS, FbZ, CkD, and CkB are the serum electrolyte level index, endocrine index, serum electrolyte level reference index, and endocrine reference index of the patient to be detected in sequence. θ1, θ2, and θ3 are the electrolyte adjustment coefficient, endocrine adjustment coefficient, and physiological regulation superposition coefficient stored in the database in sequence. ShF is the tumor physiological load index of the patient to be detected. ω1, ω2, and ω3 are the physiological function adjustment coefficient, physiological regulation adjustment coefficient, and physiological interaction coefficient stored in the database in sequence.

[0014] Furthermore, the specific steps for detecting and analyzing based on the tumor progression index of the patient to be detected and taking corresponding treatment measures based on the detection and analysis are as follows: Judge and analyze the tumor progression index of the patient to be detected with the preset tumor progression index range; if the tumor progression index of the patient to be detected is lower than the lower limit of the preset tumor progression index range, mark it as early-stage lung tumor and take the first treatment measure; if the tumor progression index of the patient to be detected is within the preset tumor progression index range, mark it as mid-stage lung tumor and take the second treatment measure; if the tumor progression index of the patient to be detected is higher than the upper limit of the preset tumor progression index range, mark it as late-stage lung tumor and take the third treatment measure.

[0015] A tumor detection system based on image analysis includes: a data acquisition module, an identification and analysis module, a comprehensive analysis module, and a detection and treatment module; the data acquisition module is used to acquire the lung image data of the patient to be detected and perform preprocessing; the identification and analysis module is used to input the preprocessed lung image data of the patient to be detected into a pre-trained region recognition model for identification and analysis to obtain several tumor regions of the patient to be detected and perform comprehensive analysis to obtain the tumor staging index of the patient to be detected; the comprehensive analysis module is used to simultaneously acquire the physiological data of the patient to be detected and perform data analysis to obtain the tumor physiological load index of the patient to be detected, and perform comprehensive analysis in combination with the tumor staging index to obtain the tumor progression index of the patient to be detected; the detection and treatment module is used to perform detection and analysis based on the tumor progression index of the patient to be detected and take corresponding treatment measures based on the detection and analysis.

[0016] The present invention has the following beneficial effects:

[0017] (1) The tumor detection method based on image analysis can accurately locate and segment the tumor area by using a U-shaped network to identify the tumor area in lung images. The decoder and skip connection structure of the U-shaped network ensure that the detailed information of the tumor area is retained, thereby improving the segmentation accuracy of the tumor boundary. For example, in the lung CT images of patients, the tumor area usually has a small difference from the surrounding tissues. The U-shaped network can provide a more refined segmentation at the tumor boundary by gradually restoring the image resolution and combining low-level features, so as to accurately detect the specific location and size of the tumor, thereby improving the accuracy of tumor detection, helping to detect small tumors or micro-lesions at an early stage, and taking intervention measures in time.

[0018] (2) The tumor detection method based on image analysis combines the image analysis results with physiological data to more comprehensively evaluate the physiological impact and progression of the tumor, thus effectively avoiding the misdiagnosis or over-treatment phenomenon that is prone to occur when relying solely on image analysis. For example, in early-stage tumors, although the image shows that the tumor is small, the physiological data may reveal the metabolic activity of the tumor, indicating that the tumor may have a risk of metastasis, so as to take targeted treatment in time.

[0019] (3) The tumor detection method based on image analysis comprehensively evaluates the tumor progression stage of the patient through the tumor progression index, so as to formulate a personalized treatment plan for the patient. For example, if the tumor progression index shows early stage, a conservative treatment plan such as surgical resection can be selected. This method can avoid the risk of delaying treatment according to the actual physiological state and progression degree of the tumor, thereby improving the survival rate of the patient.

[0020] (4) The tumor detection system based on image analysis can automatically complete the whole process from image data acquisition, processing to tumor progression assessment by integrating multiple modules, thereby improving the diagnostic efficiency, reducing the errors and omissions of manual operations, and improving the accuracy and consistency of diagnosis. Then the hospital can process more cases in a shorter time, so as to provide fast and reliable diagnostic results for patients in a busy clinical environment.

[0021] Of course, it is not necessary for any product implementing the present invention to achieve all the above advantages simultaneously. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a flowchart of a tumor detection method based on image analysis according to the present invention.

[0023] Figure 2This is the flowchart of the steps for obtaining the tumor staging index of the patient to be detected in a tumor detection method based on image analysis according to the present invention.

[0024] Figure 3 This is the block diagram of a tumor detection system based on image analysis according to the present invention. Detailed implementation manners

[0025] Please refer to Figure 1 , an embodiment of the present invention provides a technical solution: a tumor detection method based on image analysis, including the following steps: obtaining the lung image data of the patient to be detected and performing preprocessing; inputting the preprocessed lung image data of the patient to be detected into a pre-trained region recognition model for recognition and analysis to obtain several tumor regions of the patient to be detected, and performing comprehensive analysis to obtain the tumor staging index of the patient to be detected; simultaneously obtaining the physiological data of the patient to be detected and performing data analysis to obtain the tumor physiological load index of the patient to be detected, and combining with the tumor staging index for comprehensive analysis to obtain the tumor progression index of the patient to be detected; performing detection and analysis based on the tumor progression index of the patient to be detected and taking corresponding treatment measures based on the detection and analysis; wherein, the specific formula for calculating the tumor progression index of the patient to be detected is as follows: Wherein, FzL is the tumor progression index of the patient to be detected, ZqF is the tumor staging index of the patient to be detected, α1 is the staging coefficient stored in the database, β1 is the staging adjustment coefficient stored in the database, ShF is the tumor physiological load index of the patient to be detected, α2 is the load coefficient stored in the database, β2 is the load adjustment coefficient stored in the database, β3 is the interaction coefficient stored in the database, and ɑ1 + ɑ2 = 1.

[0026] It should be explained that the term exp(-β3*ZqF*ShF) in the formula is used to adjust the superposition effect of the tumor staging index and the tumor physiological load index to avoid the tumor progression index being too high or too low.

[0027] α1 and α2 can be obtained through the following steps: reading the tumor staging index and the tumor physiological load index of the patient to be detected, and performing summation analysis to obtain the progression sum value, and respectively performing ratio analysis on the tumor staging index and the tumor physiological load index of the patient to be detected with the progression sum value, and taking the ratio results as the corresponding coefficients.

[0028] β1, β2, and β3 can be obtained through the following steps: Using historical data, combined with the tumor physiological burden index and tumor staging index, perform statistical regression analysis to quantify the specific impact of each factor on the tumor progression index, thereby fitting the initial weight values. Secondly, use the sensitivity analysis method to adjust the value range of each coefficient and observe its impact on the tumor progression assessment results to ensure the stability and rationality of the model. Based on personal actual conditions, correct and optimize the initially fitted coefficients, and finally determine the coefficient values applicable to the individual.

[0029] The lung image data specifically includes the pixel values and two-dimensional coordinates of each pixel point. The physiological data includes hemoglobin concentration value, plasma lactate level value, tumor marker index, serum electrolyte level index, oxygenation index, and endocrine index.

[0030] Among them, the hemoglobin concentration value can be obtained by the sampling method. Take a blood sample from the patient to be tested, measure the sampled blood with a complete blood cell counter, and upload the results to the database. Early-stage tumors do not affect the hemoglobin level, but late-stage tumors are prone to causing anemia, resulting in a decrease in hemoglobin concentration.

[0031] The plasma lactate level value can be obtained by the sampling method. Take a blood sample from the patient to be tested, measure it with a portable blood gas analyzer, and upload the results to the database. Early-stage tumors have not yet consumed a large amount of oxygen, and the lactate level is usually close to normal. However, late-stage tumors usually have a high degree of anaerobic metabolism. Therefore, an increase in the lactate level easily reflects the active growth of tumors, especially when local hypoxia or metastasis occurs in the tumor.

[0032] The tumor marker index is the comprehensive level concentration of substances produced in tumor cells or tumor-related cells (such as carcinoembryonic antigen, carbohydrate antigen 19-9, neuron-specific enolase, etc.). It can be obtained by the sampling method. Take a blood sample from the patient to be tested, measure the concentration level of each substance with a chemiluminescence immunoassay analyzer, and perform weighted processing based on the measurement results. The obtained result is this parameter.

[0033] The serum electrolyte level index is the comprehensive concentration of electrolytes in the blood (such as sodium, potassium, calcium, chlorine, etc.). It can be obtained by the sampling method. Take a blood sample from the patient to be tested, measure the concentration level of each electrolyte with a blood gas analyzer, and perform weighted processing based on the measurement results. The obtained result is this parameter. The electrolyte level in early-stage tumors is usually relatively normal, but late-stage tumors are prone to affecting kidney function or causing tumor-related metabolic disorders, resulting in electrolyte imbalance.

[0034] The oxygenation index is used to evaluate the pulmonary gas exchange function and can be obtained through sampling. Blood samples are taken from the patient to be tested to obtain the blood oxygen saturation (the proportion of hemoglobin in the blood combined with oxygen, obtained using a pulse oximeter), the arterial partial pressure of oxygen (the actual pressure value of oxygen in arterial blood, obtained using a blood gas analyzer), and the arterial partial pressure of carbon dioxide (the partial pressure of carbon dioxide in arterial blood, obtained using a blood gas analyzer). Then, the blood oxygen saturation, arterial partial pressure of oxygen, and arterial partial pressure of carbon dioxide are standardized, and weighted processing is performed based on the standardized results. The resulting value is this parameter. Usually, the oxygenation index of early-stage tumors is normal, while in late-stage tumors, as the tumor expands or metastasizes to the lungs, the oxygenation index is likely to decrease.

[0035] The endocrine index is a comprehensive evaluation of the body's hormone levels (such as adrenaline, cortisol, insulin) and can be obtained through sampling. Blood samples are taken from the patient to be tested, and a radioimmunoassay analyzer (which uses antibodies labeled with radioactive isotopes to react with hormones in the sample and quantitatively analyze hormone levels by measuring the intensity of radioactive signals) is used to measure each hormone level. Then, weighted processing is performed based on the measurement results. The resulting value is this parameter. Early-stage tumors are difficult to affect the endocrine level, while late-stage tumors are likely to cause endocrine disorders. For example, tumors can secrete steroids, insulin, or other hormones, affecting endocrine function.

[0036] Specifically, the region recognition model is specifically a U-shaped network, and the U-shaped network includes an encoder, a bottleneck layer, a decoder, a skip connection layer, and an output layer. The specific steps to obtain several tumor regions of the patient to be detected are as follows: In the encoder of the U-shaped network, feature extraction is performed on the pixel value and two-dimensional coordinates of each pixel point of the patient to be detected (that is, the convolutional layer extracts the low-level features of the image, and the pooling layer gradually reduces the spatial size of the image. Through these operations, the encoder captures the global information of the image, and after each layer of convolution, an activation function, such as the ReLU activation function, is used for non-linear transformation to enhance the expression ability of the network), obtaining a low-resolution tumor feature map of the patient to be detected; In the bottleneck layer of the U-shaped network, compression processing is performed on the low-resolution tumor feature map of the patient to be detected (that is, after convolution operations, it will be further abstracted and compressed, enabling the network to better understand and process the global information of the input image, and the bottleneck layer usually performs several convolution and pooling operations. The size of its output feature map is relatively small, but the feature dimension is more, meaning that through these processes, the network can focus on the core information of the image, and an activation function (such as ReLU or Leaky ReLU) is used to perform non-linear conversion on the result after the convolution operation, enabling the model to process more complex image features), obtaining a high-dimensional tumor feature map of the patient to be detected; In the decoder of the U-shaped network, restoration processing is performed on the high-dimensional tumor feature map of the patient to be detected (that is, an upsampling operation is performed to interpolate the value of each pixel point, gradually expanding the spatial size of the feature map, simply increasing the resolution of the image, and then performing a transposed convolution operation, which is the opposite of the conventional convolution. The transposed convolution gradually expands the spatial dimension of the feature map through an "inverse operation" and can also retain some details of the image. Then, in each layer of the decoder, the spatial resolution of the image will be gradually restored), obtaining a restored tumor feature map of the patient to be detected (that is, a high-resolution tumor feature map); In the skip connection layer of the U-shaped network, fusion processing is performed on the restored tumor feature map of the patient to be detected (that is, during the process of restoring the image in the decoder layer, the feature map is obtained from the corresponding encoder layer and fused. The fusion method selects concatenation fusion, and the feature map of the encoder and the feature map of the decoder are concatenated along the feature dimension, so as to retain more high-resolution information in the decoding stage, which helps to improve the accuracy of segmentation, especially in the part of the tumor boundary), obtaining a fused tumor feature map of the patient to be detected (that is, retaining the detailed information extracted in the encoder, and also containing the abstract features gradually restored by the decoder, and having a higher resolution and more details);In the output layer of the U-shaped network, the fused tumor feature map of the patient to be detected is recognized (i.e., the fused tumor feature map is converted into a segmentation mask of the tumor region, achieved through pixel-level classification. The Sigmoid activation function is used in the output layer, and the output value of each pixel point will be mapped between 0 and 1, representing the probability that the pixel belongs to the tumor. After threshold segmentation, a binary image is obtained, where the pixel value of the tumor region is 1, representing the tumor region, and the background is 0, representing the non-tumor region. By performing connected component analysis on the binary image, several tumor regions of the patient to be detected can be identified and separated).

[0037] Among them, the encoder consists of a series of convolutional layers and pooling layers, whose purpose is to gradually extract the features of the image and reduce the spatial size of the image. Each layer contains two convolutional operations (usually using 3x3 convolution), and then a max pooling operation is performed. In this way, the spatial information of the image is compressed, and the feature dimension gradually increases, so as to extract deeper feature information, but the spatial resolution is relatively low.

[0038] The bottleneck layer contains convolutional and activation operations, which are used to integrate the extracted feature information and provide input for the decoding process.

[0039] The decoder gradually restores the spatial resolution of the image through upsampling (such as transposed convolution), converts the compressed feature map back to a size similar to the original image, and after each upsampling, the feature map from the corresponding layer of the encoder is fused through skip connections in order to restore the detailed information and avoid losing local features during the compression process.

[0040] The skip connection layer stitches together the high-resolution feature map of the encoder part and the feature map of the corresponding layer of the decoder part.

[0041] The output layer is usually a convolutional layer, and a 1x1 convolution is used to convert the output of the decoder into the required segmentation result.

[0042] And the pre-training process of the U-shaped network is as follows:

[0043] Obtain several groups of lung image labeled data (i.e., image data with the tumor regions in the lung images labeled), and divide the image training set and the image validation set.

[0044] Initialization: Randomly initialize the network weights, usually using the He initialization or Xavier initialization method to avoid the problem of gradient vanishing or gradient explosion during the training process.

[0045] Forward propagation: Input the image training set into the model, calculate the predicted value of each pixel, and obtain the output of the model.

[0046] Loss calculation: Calculate the loss (such as cross-entropy loss or Dice loss) between the model output and the true labels.

[0047] Backpropagation and optimization: Use the backpropagation algorithm to calculate the gradient of the loss function with respect to the model weights.

[0048] Use an optimization algorithm (such as the Adam optimizer) to update the model weights to minimize the loss function.

[0049] Iterative training: Through multiple training epochs, continuously optimize the model parameters until the model converges.

[0050] Hyperparameter tuning: Use the image validation set to evaluate the model's performance during training, and adjust hyperparameters (such as learning rate, batch size, network depth, etc.) to prevent overfitting or underfitting.

[0051] In this implementation, through the multi-layer convolution and pooling operations of the encoder, low-level to high-level features of the image can be gradually extracted. At the same time, the spatial size of the image is reduced during the pooling operation to capture global information, which helps the model better understand complex image content, especially the different shapes and boundaries of tumors. Through multiple convolutions and poolings, the bottleneck layer enables the network to extract more abstract and high-dimensional features, thereby enhancing the processing ability for complex images. Secondly, the design of the skip connection layer enables the decoder to obtain more high-resolution detail information from the encoder, effectively avoiding the loss of important local features during compression, and thus contributing to improving the segmentation accuracy, especially for the boundary recognition of tumors. Finally, the output layer uses the Sigmoid activation function, and a binary image of the tumor area is obtained through threshold segmentation. After connected component analysis, multiple tumor areas can be separated, and pixel-level classification is performed to accurately identify the location and shape of the tumors.

[0052] Specifically, as Figure 2As shown below, the specific steps to obtain the tumor staging index of the patient to be detected are as follows: Read the two-dimensional coordinates of each pixel point in each tumor region of the patient to be detected, and perform median analysis to obtain the two-dimensional coordinates of each tumor region of the patient to be detected, and perform comprehensive analysis (that is, analyze several groups of adjacent tumor region distances through the Euclidean distance formula and perform standard deviation processing) to obtain the tumor distribution index of the patient to be detected; perform comprehensive analysis on the two-dimensional coordinates of each pixel point in each tumor region of the patient to be detected to obtain the smoothness index of each tumor region of the patient to be detected; and perform gray-scale processing on the pixel values of each pixel point in each tumor region of the patient to be detected to obtain the gray-scale pixel value of each pixel point in each tumor region of the patient to be detected, and perform comprehensive analysis with the gray-scale pixel values of each neighborhood pixel point in the set neighborhood respectively to obtain the texture index of each tumor region of the patient to be detected; and perform comprehensive analysis on the tumor distribution index of the patient to be detected and the smoothness index and texture index of each tumor region to obtain the tumor staging index of the patient to be detected.

[0053] The specific formulas for calculating the texture index and tumor staging index of each tumor region of the patient to be detected are as follows: Among them, WZS i is the texture index of the i-th tumor region of the patient to be detected, Xa ijm is the gray-scale pixel value of the m-th neighborhood pixel point in the set neighborhood of the j-th pixel point in the i-th tumor region of the patient to be detected, Xa ij is the gray-scale pixel value of the j-th pixel point in the i-th tumor region of the patient to be detected, S(X) is the sign function, ZqF is the tumor staging index of the patient to be detected, QbF is the tumor distribution index of the patient to be detected, δ1 is the distribution adjustment coefficient stored in the database, PhS i is the smoothness index of the i-th tumor region of the patient to be detected, δ2 is the smoothness adjustment coefficient stored in the database, δ3 is the texture adjustment coefficient stored in the database, δ4 is the superposition coefficient stored in the database, i = 1, 2, 3,..., i0, i0 is the number of tumor regions, j = 1, 2, 3,..., j0, j0 is the number of pixel points, m = 1, 2, 3,..., m0, m0 is the number of neighborhood pixel points, e is the natural constant, and in this embodiment, its value is 2.71.

[0054] It should be explained that δ1, δ2, δ3, and δ can be obtained through the following steps: Based on historical data, determine the initial influence weights of each variable (such as tumor distribution index, smoothing index, texture index) on the tumor staging index through statistical regression analysis. Then, use the sensitivity analysis method to adjust the value range of the coefficients to evaluate the stability and applicability of these parameters to the formula output. Next, further fit the weights through model optimization (such as machine learning algorithms or multi-objective optimization) to ensure that the formula can accurately reflect the actual tumor staging status. Fine-tune the coefficients based on different population characteristics to ensure its applicability to specific tumor staging assessment requirements.

[0055] In this implementation plan, by performing median analysis, distance analysis, and grayscale processing on each pixel point of each tumor region, the spatial distribution, morphological characteristics, and texture characteristics of the tumor region can be described in detail, which helps to improve the accuracy of the tumor staging index, thereby ensuring that the staging result can fully reflect the actual state of the tumor. Secondly, by analyzing the adjacent distance between tumor regions through the Euclidean distance formula and performing standard deviation processing, the distribution characteristics of the tumor can be better evaluated, thereby accurately identifying the spread range of the tumor and its relationship with the surrounding tissues, and then improving the accuracy of tumor staging. Moreover, the smoothing index and texture index provide in-depth analysis of the morphology and internal structure of the tumor region, thereby revealing the complexity and heterogeneity of the internal tissues of the tumor, which helps to identify the malignancy and growth characteristics of the tumor. At the same time, through comprehensive analysis of multiple indicators such as tumor distribution index, smoothing index, and texture index, it helps to comprehensively understand the characteristics of the tumor, thereby providing a more comprehensive and accurate staging assessment. Finally, through statistical regression analysis and sensitivity analysis based on historical data, the value range of each coefficient can be reasonably adjusted, and the reliability of the staging index can be further improved through model optimization, thus ensuring the effectiveness and accuracy of the tumor staging index, and then being able to adapt to the changes of different patients and tumors.

[0056] Specifically, the specific steps to obtain the smoothness index of each tumor region of the patient to be detected are as follows: Read the two-dimensional coordinates of each boundary pixel point of each tumor region of the patient to be detected, and perform comprehensive analysis in sequence to obtain the curvature index and boundary uniformity index of each tumor region of the patient to be detected (comprehensive analysis means analyzing the distances of several groups of adjacent boundary pixel points through the Euclidean distance formula and performing standard deviation processing); Read each pixel point of each tumor region of the patient to be detected and perform comprehensive analysis (i.e., statistical analysis, counting the total number of pixel points in each tumor region, which is the area) to obtain the actual area value of each tumor region of the patient to be detected; And perform convex hull analysis on the two-dimensional coordinates of each boundary pixel point of each tumor region of the patient to be detected (i.e., based on the convex hull algorithm, obtain a convex polygon containing the boundary points and perform calculation processing based on the polygon area formula) to obtain the convex hull area value of each tumor region of the patient to be detected, and perform comprehensive analysis in combination with the actual area value (i.e., ratio analysis, actual area value / convex hull area value) to obtain the compactness index of each tumor region of the patient to be detected; And perform comprehensive analysis on the curvature index, boundary uniformity index, and compactness index of each tumor region of the patient to be detected to obtain the smoothness index of each tumor region of the patient to be detected.

[0057] Among them, the specific formulas for calculating the curvature index and smoothness index of each tumor region of the patient to be detected are as follows: Among them, QzL i is the curvature index of the i-th tumor region of the patient to be detected, (X in , Y in ) is the two-dimensional coordinate of the n-th boundary pixel point of the i-th tumor region of the patient to be detected, (X i(n+1) , Y i(n+1) ) is the two-dimensional coordinate of the (n + 1)-th boundary pixel point of the i-th tumor region of the patient to be detected, (X i(n-1) , Y i(n-1) ) is the two-dimensional coordinate of the (n - 1)-th boundary pixel point of the i-th tumor region of the patient to be detected, PhS i is the smoothness index of the i-th tumor region of the patient to be detected, η1 is the curvature coefficient stored in the database, ByZ i is the boundary uniformity index of the i-th tumor region of the patient to be detected, η2 is the boundary uniformity coefficient stored in the database, CdZ i is the compactness index of the i-th tumor region of the patient to be detected, η3 is the compactness coefficient stored in the database, η1 + η2 + η3 = 1, i = 1, 2, 3,..., i0, i0 is the number of tumor regions, n = 1, 2, 3..., n0, n0 is the number of boundary pixel points.

[0058] It should be explained that for i = 1, use (X i1 , Y i1), (X i2 , Y i2 ), Calculate using three coordinates. And for i = n0, use (x i1 , Y i1 ), Calculate using three coordinates.

[0059] η1, η2, and η3 can be obtained through the following steps: Read the curvature index, boundary uniformity index, and compactness index of each tumor region of the patient to be detected, perform mean analysis to obtain the mean curvature index, mean boundary uniformity index, and mean compactness index of the patient to be detected, and perform summation analysis to obtain the smooth sum value. Then, perform ratio analysis on the mean curvature index, mean boundary uniformity index, and mean compactness index of the patient to be detected with the smooth sum value respectively, and use the ratio results as the corresponding coefficients.

[0060] In this implementation scheme, by comprehensively analyzing the curvature index and boundary uniformity index of the boundary pixel points of the tumor region, the smoothness and regularity of the tumor boundary can be accurately evaluated. By using the Euclidean distance formula to calculate the distance between adjacent boundary pixel points and performing standard deviation processing, it helps to precisely evaluate the smoothness and uniformity of the tumor boundary, thereby enabling the quantification of the shape characteristics of the tumor region, avoiding the errors caused by pure visual evaluation, and providing more reliable data support. Secondly, by performing convex hull analysis on the tumor boundary, the convex hull area of the tumor can be obtained, and ratio analysis is combined with the actual area value to effectively evaluate the compactness of the tumor region. The compactness index can reflect the regularity of the tumor morphology. If the compactness is high, it indicates that the tumor is relatively regular; otherwise, there may be irregular expansion or other pathological characteristics. Finally, by comprehensively analyzing the curvature index, boundary uniformity index, and compactness index, the smoothness of the tumor region can be comprehensively evaluated from multiple dimensions, providing more comprehensive tumor morphology information, which helps to understand the growth characteristics, expansion mode, and relationship with surrounding tissues of the tumor.

[0061] Specifically, the specific steps to obtain the tumor physiological load index of the patient to be tested are as follows: Obtain the reference value of hemoglobin concentration, reference value of plasma lactate level, reference index of tumor markers, reference index of serum electrolyte level, reference index of oxygenation, and reference index of endocrine for the patient to be tested; comprehensively analyze the reference value of hemoglobin concentration, reference value of plasma lactate level, reference index of tumor markers, reference index of oxygenation, hemoglobin concentration value, plasma lactate level value, tumor marker index, and oxygenation index of the patient to be tested to obtain the tumor physiological function impact index of the patient to be tested (i.e., the physiological function of the human body at different tumor stages); and comprehensively analyze the reference index of serum electrolyte level, reference index of endocrine, serum electrolyte level index, and endocrine index of the patient to be tested to obtain the physiological regulation index of the patient to be tested (i.e., the physiological regulation ability of the human body at different tumor stages).

[0062] And comprehensively analyze the tumor physiological function impact index and physiological regulation index of the patient to be tested to obtain the tumor physiological load index of the patient to be tested.

[0063] The specific formulas for calculating the tumor physiological function impact index, physiological regulation index, and tumor physiological load index of the patient to be tested are as follows: Wherein, LsG is the tumor physiological function impact index of the patient to be tested, CkH is the reference value of hemoglobin concentration of the patient to be tested, XhD is the hemoglobin concentration value of the patient to be tested, μ1 is the hemoglobin adjustment coefficient stored in the database, RsP is the plasma lactate level value of the patient to be tested, CsP is the reference value of plasma lactate level of the patient to be tested, μ2 is the lactate adjustment coefficient stored in the database, ZbW is the tumor marker index of the patient to be tested, CbW is the reference index of tumor markers of the patient to be tested, μ3 is the marker adjustment coefficient stored in the database, CkY is the reference index of oxygenation of the patient to be tested, YhZ is the oxygenation index of the patient to be tested, μ4 is the oxygenation adjustment coefficient stored in the database, μ5 is the functional impact superposition coefficient stored in the database, TsL is the physiological regulation index of the patient to be tested, XdS is the serum electrolyte level index of the patient to be tested, CkD is the reference index of serum electrolyte level of the patient to be tested, θ1 is the electrolyte adjustment coefficient stored in the database, FbZ is the endocrine index of the patient to be tested, CkB is the reference index of endocrine of the patient to be tested, θ2 is the endocrine adjustment coefficient stored in the database, θ3 is the physiological regulation superposition coefficient stored in the database, ShF is the tumor physiological load index of the patient to be tested, ω1 is the physiological function adjustment coefficient stored in the database, ω2 is the physiological regulation adjustment coefficient stored in the database, and ω3 is the physiological interaction coefficient stored in the database.

[0064] It should be noted that μ1, μ2, μ3, μ4, and μ5 can be obtained through the following steps: Using historical data, evaluate the influence degree of each variable (such as hemoglobin concentration value, plasma lactate level value, etc.) on the tumor physiological function impact index through statistical modeling and regression analysis, so as to fit the initial weight value. Then, based on sensitivity analysis, adjust the value range of these coefficients to ensure that the formula has good adaptability to the changes in tumor physiological function under different populations, and reasonably correct the weight coefficients according to the specific population characteristics.

[0065] θ1, θ2, and θ3 can be obtained through the following steps: Using historical data, evaluate the influence degree of each variable (such as serum electrolyte level index, endocrine index, etc.) on the physiological regulation index through statistical modeling and regression analysis, so as to fit the initial weight value. Then, based on sensitivity analysis, adjust the value range of these coefficients to ensure that the formula has good adaptability to the physiological regulation changes under different populations, and reasonably correct the weight coefficients according to the specific population characteristics.

[0066] ω1, ω2, and ω3 can be obtained through the following steps: Using historical data, combining the tumor physiological function impact index and the physiological regulation index, conduct statistical regression analysis to quantify the specific influence of each factor on the tumor physiological load index, so as to fit the initial weight value. Secondly, adopt the sensitivity analysis method to adjust the value range of each coefficient and observe its influence on the tumor physiological load assessment result to ensure the stability and rationality of the model. Based on the individual's actual situation, correct and optimize the initially fitted coefficients, and finally determine the coefficient values applicable to the individual.

[0067] The specific implementation example of calculating the tumor physiological load index of the patient to be detected is as follows. The following data is available: including hemoglobin concentration value, plasma lactate level value, tumor marker index, serum electrolyte level index, oxygenation index, and endocrine index. The specific data is shown in Tables 1 and 2:

[0068] Table 1 Example of Physiological Data of the Patient to be Detected

[0069]

[0070] Table 2 Example of Physiological Reference Data of the Patient to be Detected

[0071]

[0072]

[0073] The hemoglobin adjustment coefficient μ1 stored in the database is approximately: 0.36;

[0074] The lactate adjustment coefficient μ2 stored in the database is approximately: 0.42;

[0075] The marker adjustment coefficient μ3 stored in the database is approximately: 0.29;

[0076] The oxygenation adjustment coefficient μ4 stored in the database is approximately: 0.51;

[0077] The functional impact superposition coefficient μ5 stored in the database is approximately: 0.69;

[0078] The electrolyte adjustment coefficient θ1 stored in the database is approximately: 0.23;

[0079] The endocrine adjustment coefficient θ2 stored in the database is approximately: 0.19;

[0080] The physiological regulation superposition coefficient θ3 stored in the database is approximately: 0.14;

[0081] The physiological function adjustment coefficient ω1 stored in the database is approximately: 0.48;

[0082] The physiological regulation adjustment coefficient ω2 stored in the database is approximately: 0.61;

[0083] The physiological interaction coefficient ω3 stored in the database is approximately: 1.46;

[0084] Substitute the above coefficients and the data in Table 1 and Table 2 into the specific formulas for calculating the tumor physiological function impact index, physiological regulation index, and tumor physiological load index of the patient to be tested, and we get:

[0085] The tumor physiological function impact index of the patient to be tested = exp((15.50 / 17.32) 0.36 * (1.48 / 1.35) 0.42 * (19.86 / 18.17) 0.29 * (0.73 / 0.43) 0.51 ) 0.69 ≈ 3.82.

[0086] The physiological regulation index of the patient to be tested = ln(1 + ((|76.49 - 68.76| / 68.76) 0.23 * (|18.36 - 23.84| / 23.84) 0.19 ) 0.14 ) ≈ 0.64.

[0087] The tumor physiological load index of the patient to be tested = exp(0.48 * 3.82 * (1 + 0.61 * √0.64)) / 1.46 * 3.82 * 0.64 ≈ 2.13.

[0088] In this implementation, by combining multiple physiological parameters, the physiological state of the patient to be detected is comprehensively evaluated, thereby reflecting the comprehensive burden of the tumor on the patient's physiological function and physiological regulation ability. Through regression analysis, sensitivity analysis, and model optimization, and by adjusting the coefficients in combination with the specific data and personalized characteristics of different patients, the calculation results can be adapted to the differences of different populations, thus ensuring a more accurate assessment of physiological load. Secondly, by combining the tumor physiological function impact index and the physiological regulation index and comprehensively analyzing the relationship between the two, the interaction effects between these factors can be considered, so as to more truly reflect the impact of the tumor on the overall physiological load of the patient, thereby providing a more comprehensive and accurate assessment of physiological load. Finally, by dynamically adjusting and optimizing the coefficients, this method is not only applicable to a specific population, but can also be fine-tuned according to different clinical needs. Moreover, the physical states and disease development stages of different patients may vary, and personalized adjustment ensures a more accurate assessment.

[0089] Specifically, the specific steps for detecting and analyzing based on the tumor progression index of the patient to be detected and taking corresponding treatment measures are as follows: Judge and analyze the tumor progression index of the patient to be detected with the preset tumor progression index interval; If the tumor progression index of the patient to be detected is lower than the lower limit of the preset tumor progression index interval (i.e., the minimum value of the tumor progression index interval), it is marked as early-stage lung tumor, and the first treatment measure is taken (i.e., providing suggestions of surgical resection to relevant personnel. If the tumor is very small, minimally invasive surgery or thoracoscopic surgery can be considered to reduce the trauma to the body, radiotherapy. For patients who cannot undergo surgery, radiotherapy can be selected to directly kill tumor cells with high-energy rays, stereotactic radiotherapy, concentrating high-dose radiation to kill tumors and reduce the harm to normal tissues); If the tumor progression index of the patient to be detected is within the preset tumor progression index interval, it is marked as mid-stage lung tumor, and the second treatment measure is taken (i.e., providing suggestions of chemotherapy to relevant personnel, adjuvant chemotherapy to control tumor spread, chemotherapy drugs such as cisplatin, gemcitabine, pemetrexed; targeted therapy. For patients with non-small cell lung cancer with specific gene mutations, targeted drug therapy can be used, and common targeted drugs include erlotinib, afatinib, etc.); If the tumor progression index of the patient to be detected is higher than the upper limit of the preset tumor progression index interval (i.e., the maximum value of the tumor progression index interval), it is marked as late-stage lung tumor, and the third treatment measure is taken (i.e., providing suggestions of immunotherapy to relevant personnel, by inhibiting the tumor immune escape mechanism and activating the body's immune system to attack tumor cells, common drugs include nivolumab, pembrolizumab, etc.; palliative care, that is, relieving symptoms and improving the quality of life, specifically including measures such as pain relief, respiratory support, and relieving loss of appetite).

[0090] In this embodiment, through the quantitative analysis of the tumor progression index, the tumor progression of patients is divided into early, middle, and late stages, which can help doctors formulate personalized treatment plans for each patient, thereby contributing to improving the treatment effect and cure rate. Different treatment methods are adopted for patients at different stages, which helps to formulate personalized treatment plans, accurately select the most suitable treatment method for patients, and improve the quality of life of patients. At the same time, as the patient's treatment process progresses, the change in the tumor progression index can help doctors dynamically adjust the treatment plan.

[0091] Please refer to Figure 3 , an embodiment of the present invention provides a technical solution: a tumor detection system based on image analysis, including: a data acquisition module, an identification and analysis module, a comprehensive analysis module, and a detection and treatment module; the data acquisition module is used to acquire the lung image data of the patient to be detected and perform preprocessing; the identification and analysis module is used to input the preprocessed lung image data of the patient to be detected into a pre-trained region identification model for identification and analysis, obtain several tumor regions of the patient to be detected, and perform comprehensive analysis to obtain the tumor staging index of the patient to be detected; the comprehensive analysis module is used to simultaneously acquire the physiological data of the patient to be detected, perform data analysis to obtain the tumor physiological load index of the patient to be detected, and perform comprehensive analysis in combination with the tumor staging index to obtain the tumor progression index of the patient to be detected; the detection and treatment module is used to perform detection and analysis based on the tumor progression index of the patient to be detected and take corresponding treatment measures based on the detection and analysis.

[0092] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they know the basic creative concept. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications falling within the scope of the present invention.

[0093] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. A tumor detection method based on image analysis, characterized in that, It includes the following steps: Obtain the lung image data of the patient to be detected and perform preprocessing; Input the preprocessed lung image data of the patient to be detected into a pre-trained region recognition model for recognition and analysis, obtain several tumor regions of the patient to be detected, and perform comprehensive analysis to obtain the tumor staging index of the patient to be detected; At the same time, obtain the physiological data of the patient to be detected and perform data analysis to obtain the tumor physiological load index of the patient to be detected, and combine it with the tumor staging index for comprehensive analysis to obtain the tumor progression index of the patient to be detected; Perform detection and analysis based on the tumor progression index of the patient to be detected, and take corresponding treatment measures based on the detection and analysis; Among them, the specific formula for calculating the tumor progression index of the patient to be detected is as follows: Among them, F zL , Z qF , S hF are, in sequence, the tumor progression index, tumor staging index, and tumor physiological burden index of the patient to be detected, and α1, β1, α2, β2, β3 are, in sequence, the staging coefficient, staging adjustment coefficient, burden coefficient, burden adjustment coefficient, and interaction coefficient stored in the database, and α1 + α2 = 1.

2. The tumor detection method based on image analysis according to claim 1, wherein The lung image data is specifically the pixel value and two-dimensional coordinates of each pixel point, and the physiological data includes hemoglobin concentration value, plasma lactate level value, tumor marker index, serum electrolyte level index, oxygenation index, and endocrine index.

3. The tumor detection method based on image analysis according to claim 2, characterized in that, The region recognition model is specifically a U-shaped network, and the U-shaped network includes an encoder, a bottleneck layer, a decoder, a skip connection layer, and an output layer. The specific steps for obtaining several tumor regions of the patient to be detected are as follows: In the encoder of the U-shaped network, extract features from the pixel value and two-dimensional coordinates of each pixel point of the patient to be detected to obtain a low-resolution tumor feature map of the patient to be detected; In the bottleneck layer of the U-shaped network, perform compression processing on the low-resolution tumor feature map of the patient to be detected to obtain a high-dimensional tumor feature map of the patient to be detected; In the decoder of the U-shaped network, perform restoration processing on the high-dimensional tumor feature map of the patient to be detected to obtain a restored tumor feature map of the patient to be detected; In the skip connection layer of the U-shaped network, perform fusion processing on the restored tumor feature map of the patient to be detected to obtain a fused tumor feature map of the patient to be detected; In the output layer of the U-shaped network, perform recognition processing on the fused tumor feature map of the patient to be detected to obtain several tumor regions of the patient to be detected.

4. The tumor detection method based on image analysis according to claim 3, wherein The specific steps for obtaining the tumor staging index of the patient to be detected are as follows: Read the two-dimensional coordinates of each pixel point of each tumor region of the patient to be detected and perform median analysis to obtain the two-dimensional coordinates of each tumor region of the patient to be detected, and perform comprehensive analysis to obtain the tumor distribution index of the patient to be detected; Perform comprehensive analysis on the two-dimensional coordinates of each pixel point of each tumor region of the patient to be detected to obtain the smoothness index of each tumor region of the patient to be detected; And perform gray processing on the pixel values of each pixel point of each tumor region of the patient to be detected to obtain the gray pixel values of each pixel point of each tumor region of the patient to be detected, and perform comprehensive analysis with the gray pixel values of each neighborhood pixel point in the set neighborhood respectively to obtain the texture index of each tumor region of the patient to be detected; And perform comprehensive analysis on the tumor distribution index of the patient to be detected and the smoothness index and texture index of each tumor region to obtain the tumor staging index of the patient to be detected.

5. The tumor detection method based on image analysis according to claim 4, characterized in that, The specific formulas for calculating the texture index of each tumor region and the tumor staging index of the patient to be detected are as follows: Among them, WzS i is the texture index of the i-th tumor region of the patient to be detected, Xa ijm is the gray pixel value of the m-th neighborhood pixel of the set neighborhood of the j-th pixel of the i-th tumor region of the patient to be detected, Xa ij is the gray pixel value of the j-th pixel of the i-th tumor region of the patient to be detected, S(X) is the sign function, ZqF and QbF are the tumor stage index and tumor distribution index of the patient to be detected in sequence, PhS i is the smoothing index of the i-th tumor region of the patient to be detected, δ1, δ2, δ3, δ4 are the distribution adjustment coefficient, smoothing adjustment coefficient, texture adjustment coefficient, and superposition coefficient stored in the database in sequence, i = 1, 2, 3, …, i0, i0 is the number of tumor regions, j = 1, 2, 3, …, j0, j0 is the number of pixel points, m = 1, 2, 3, …, m0, m0 is the number of neighborhood pixel points, and e is the natural constant.

6. The tumor detection method based on image analysis according to claim 4, wherein The specific steps to obtain the smoothness index of each tumor region of the patient to be detected are as follows: Read the two-dimensional coordinates of each boundary pixel point of each tumor region of the patient to be detected, and conduct comprehensive analysis in sequence to obtain the curvature index and boundary uniformity index of each tumor region of the patient to be detected; Read each pixel point of each tumor region of the patient to be detected, and conduct comprehensive analysis to obtain the actual area value of each tumor region of the patient to be detected; Conduct convex hull analysis on the two-dimensional coordinates of each boundary pixel point of each tumor region of the patient to be detected to obtain the convex hull area value of each tumor region of the patient to be detected, and conduct comprehensive analysis in combination with the actual area value to obtain the compactness index of each tumor region of the patient to be detected; Conduct comprehensive analysis on the curvature index, boundary uniformity index, and compactness index of each tumor region of the patient to be detected to obtain the smoothness index of each tumor region of the patient to be detected.

7. The tumor detection method based on image analysis according to claim 2, wherein The specific steps to obtain the tumor physiological load index of the patient to be detected are as follows: Obtain the reference value of hemoglobin concentration, reference value of plasma lactate level, reference index of tumor markers, reference index of serum electrolyte level, reference index of oxygenation, and reference index of endocrine of the patient to be detected; Conduct comprehensive analysis on the reference value of hemoglobin concentration, reference value of plasma lactate level, reference index of tumor markers, reference index of oxygenation, hemoglobin concentration value, plasma lactate level value, tumor marker index, and oxygenation index of the patient to be detected to obtain the tumor physiological function influence index of the patient to be detected; Conduct comprehensive analysis on the reference index of serum electrolyte level, reference index of endocrine, serum electrolyte level index, and endocrine index of the patient to be detected to obtain the physiological regulation index of the patient to be detected; Conduct comprehensive analysis on the tumor physiological function influence index and physiological regulation index of the patient to be detected to obtain the tumor physiological load index of the patient to be detected.

8. The tumor detection method based on image analysis according to claim 7, characterized in that, The specific formulas for calculating the tumor physiological function influence index, physiological regulation index, and tumor physiological load index of the patient to be detected are as follows: Among them, LsG is the tumor physiological function impact index of the patient to be detected. CkH, CsP, CbW, and CkY are the reference values of hemoglobin concentration, reference value of plasma lactate level, reference index of tumor markers, and reference index of oxygenation of the patient to be detected in sequence. XhD, RsP, ZbW, and YhZ are the hemoglobin concentration value, plasma lactate level value, tumor marker index, and oxygenation index of the patient to be detected in sequence. μ1, μ2, μ3, μ4, and μ5 are the hemoglobin adjustment coefficient, lactate adjustment coefficient, marker adjustment coefficient, oxygenation adjustment coefficient, and function impact superposition coefficient stored in the database in sequence. TsL is the physiological regulation index of the patient to be detected. XdS, FbZ, CkD, and CkB are the serum electrolyte level index, endocrine index, serum electrolyte level reference index, and endocrine reference index of the patient to be detected in sequence. θ1, θ2, and θ3 are the electrolyte adjustment coefficient, endocrine adjustment coefficient, and physiological regulation superposition coefficient stored in the database in sequence. ShF is the tumor physiological load index of the patient to be detected. ω1, ω2, and ω3 are the physiological function adjustment coefficient, physiological regulation adjustment coefficient, and physiological interaction coefficient stored in the database in sequence.

9. The tumor detection method based on image analysis according to claim 1, characterized in that The specific steps for detecting and analyzing based on the tumor progression index of the patient to be detected and taking corresponding treatment measures based on the detection and analysis are as follows: Judge and analyze the tumor progression index of the patient to be detected with the preset tumor progression index range; If the tumor progression index of the patient to be detected is lower than the lower limit of the preset tumor progression index range, it is marked as early-stage lung tumor, and the first treatment measure is taken; If the tumor progression index of the patient to be detected is within the preset tumor progression index range, it is marked as mid-stage lung tumor, and the second treatment measure is taken; If the tumor progression index of the patient to be detected is higher than the upper limit of the preset tumor progression index range, it is marked as late-stage lung tumor, and the third treatment measure is taken.

10. A tumor detection system based on image analysis, which applies the tumor detection method based on image analysis according to any one of claims 1-9, is characterized in that, Including: Data acquisition module, recognition and analysis module, comprehensive analysis module, detection and treatment module; The data acquisition module is used to acquire the lung image data of the patient to be detected and perform preprocessing; The recognition and analysis module is used to input the preprocessed lung image data of the patient to be detected into a pre-trained region recognition model for recognition and analysis, obtain several tumor regions of the patient to be detected, and perform comprehensive analysis to obtain the tumor staging index of the patient to be detected; The comprehensive analysis module is used to simultaneously acquire the physiological data of the patient to be detected, perform data analysis to obtain the tumor physiological load index of the patient to be detected, and perform comprehensive analysis in combination with the tumor staging index to obtain the tumor progression index of the patient to be detected; The detection and treatment module is used to perform detection and analysis based on the tumor progression index of the patient to be detected and take corresponding treatment measures based on the detection and analysis.

Images